CN105378948B

CN105378948B - Dicing a wafer of light emitting devices

Info

Publication number: CN105378948B
Application number: CN201480041215.6A
Authority: CN
Inventors: S.R.佩达达; F.L.魏
Original assignee: Bright Sharp Holdings Ltd
Current assignee: Lumileds Holding BV
Priority date: 2013-07-18
Filing date: 2014-07-07
Publication date: 2020-08-28
Anticipated expiration: 2034-07-07
Also published as: TW201511323A; US20160163934A1; US10707387B2; TWI713232B; EP3118904A1; TWI743823B; EP3118903B1; TW202215677A; EP3022777B1; EP3118903A1; KR20160032221A; JP2016525286A; CN105378948A; EP3118904B1; JP6568062B2; EP3022777A2; WO2015008189A3; WO2015008189A2

Abstract

Some embodiments include a ill-nitride light emitting device with a light emitting layer disposed between an n-type region and a p-type region. The glass layer is connected to the III-nitride light emitting device. A wavelength converting layer is disposed between the III-nitride light emitting device and the glass layer. The glass layer is narrower than the ill-nitride light emitting device.

Description

Dicing a wafer of light emitting devices

Technical Field

The invention relates to dicing a wafer of light emitting devices.

Background

Semiconductor light emitting devices, including Light Emitting Diodes (LEDs), Resonant Cavity Light Emitting Diodes (RCLEDs), vertical cavity laser diodes (VCSELs), and edge emitting lasers, are the most efficient light sources currently available. Materials systems currently of interest in the fabrication of high-brightness light emitting devices capable of operation across the visible spectrum include group III-V semiconductors, particularly binary, ternary, and quaternary alloys of gallium, aluminum, indium, and nitrogen (also referred to as ill-nitride materials). In general, III-nitride light emitting devices are fabricated by epitaxially growing a stack of semiconductor layers of different composition and doping concentration on a sapphire, silicon carbide, III-nitride, or other suitable substrate via Metal Organic Chemical Vapor Deposition (MOCVD), Molecular Beam Epitaxy (MBE), or other epitaxial techniques. The stack typically includes one or more n-type layers doped with, for example, Si, formed over a substrate, one or more light emitting layers in an active region formed over the n-type layer(s), and one or more p-type layers doped with, for example, Mg, formed over the active region. Electrical contacts are formed on the n-and p-type regions.

Disclosure of Invention

It is an object of the present invention to provide a method of dicing a wafer comprising a light emitting device, a wavelength converting layer and a transparent layer.

Some embodiments include a method of dicing a wafer of light emitting devices. The wafer includes a glass layer, a light emitting device layer comprising a plurality of light emitting devices separated by a dielectric, and a wavelength converting layer disposed between the glass layer and the light emitting device layer. The method includes sawing the wafer in the dielectric region with a metal bonded diamond grit blade.

Some embodiments include a method of dicing a wafer of light emitting devices. The wafer includes a transparent layer, a light emitting device layer including a plurality of light emitting devices separated by a dielectric, and a wavelength converting layer disposed between the transparent layer and the light emitting device layer. The method includes cutting a first portion of a thickness of the wafer in a first cutting process and cutting the remaining thickness of the wafer in a second cutting process.

Some embodiments include a ill-nitride light emitting device with a light emitting layer disposed between an n-type region and a p-type region. The glass layer is connected to the III-nitride light emitting device. A wavelength converting layer is disposed between the III-nitride light emitting device and the glass layer. The glass layer has a smaller lateral extent in a plane parallel to a top surface of the ill-nitride light emitting device than the ill-nitride light emitting device.

Drawings

Fig. 1 illustrates a wafer comprising LEDs, a wavelength converting layer and a transparent layer.

Fig. 2 illustrates an example of an LED.

Fig. 3 illustrates the wafer of fig. 1 arranged on a dicing frame.

Fig. 4A illustrates partial sawing of a wafer with a wide blade. Fig. 4B illustrates sawing through the remaining thickness of the wafer in fig. 4A with a narrow blade.

Fig. 5A illustrates partial sawing of a wafer with a wide blade. Fig. 5B illustrates cutting through the remaining thickness of the wafer in fig. 5A using a lift-off laser.

Fig. 6A illustrates partial sawing of a wafer with a narrow blade. Fig. 6B illustrates flipping the wafer upside down and sawing through the remaining thickness of the wafer in fig. 6A with a wide blade.

Fig. 7A illustrates partially dicing a wafer using a lift-off laser. Fig. 7B illustrates flipping the wafer upside down and sawing through the remaining thickness of the wafer in fig. 7A with a wide blade.

Fig. 8A illustrates partially dicing a wafer using a lift-off laser. Fig. 8B illustrates flipping the wafer upside down and scratching and breaking the remaining thickness of the wafer in fig. 8A.

Fig. 9A illustrates partial sawing of a wafer with a narrow blade. Fig. 9B illustrates flipping the wafer upside down and scratching and breaking the remaining thickness of the wafer in fig. 9A.

Detailed Description

Although in the following examples the semiconductor light emitting device is a ill-nitride LED emitting blue or UV light, semiconductor light emitting devices other than LEDs may be used, such as laser diodes and semiconductor light emitting devices made of other material systems, such as other III-V materials, III phosphides, III arsenides, II-VI materials, ZnO or silicon based materials.

Fig. 1 illustrates a portion of a wafer 100 of III-nitride LEDs. A plurality of LEDs 10 are attached to the wavelength converting layer 14 and the transparent layer 16. In some embodiments, wavelength converting layer 14 is disposed between transparent layer 16 and LED 10.

The wavelength converting layer 14 may be, for example, wavelength converting particles, such as a powder phosphor, disposed in a transparent material such as silicone. The wavelength-converting layer 14 may be a flexible film that is formed separately from the LED10 and then laminated over the wafer of LEDs 10.

Transparent layer 16 may be, for example, a glass layer. In some embodiments, the transparent layer 16 may comprise other non-transparent materials, such as scattering particles or wavelength converting particles.

A dielectric material (shaded portion) 12, such as epoxy, separates adjacent LEDs 10. Other materials such as reflective particles may be disposed in the dielectric material.

Fig. 2 illustrates one example of a single LED 10. Any suitable semiconductor light emitting device may be used and embodiments of the invention are not limited to the device illustrated in fig. 2. The LED illustrated in fig. 2 and the portion of the wafer 100 illustrated in fig. 1 may be formed as follows. Semiconductor structure 22 is grown on growth substrate 20 as is known in the art. The growth substrate is typically sapphire, but may be any suitable substrate, such as, for example, SiC, Si, GaN, or composite substrates. Semiconductor structure 22 includes a light emitting or active region sandwiched between n and p-type regions. n-type region 24 may be grown first and may include multiple layers of different compositions and doping concentrations, including, for example, preparation layers (such as buffer layers or nucleation layers) and/or layers designed to facilitate removal of the growth substrate (which may be n-type or not intentionally doped) and n-or even p-type device layers designed for specific optical, material, or electrical properties desired for efficient light emission from the light emitting region. A light emitting or active region 26 is grown over the n-type region. Examples of suitable light emitting regions include a single thick or thin light emitting layer, or multiple quantum well light emitting regions containing multiple thin or thick light emitting layers separated by barrier layers. A p-type region 28 may then be grown over the light emitting region. Like n-type region 24, p-type region 28 may include multiple layers of different composition, thickness, and doping concentration, including layers that are not intentionally doped or n-type layers.

After growth of the semiconductor structure, a p-contact 30 is formed on the surface of p-type region 28. The p-contact 30 typically includes multiple conductive layers, such as a reflective metal and a shield metal that can prevent or reduce electromigration of the reflective metal. The reflective metal is typically silver, but any suitable material or materials may be used. After forming the p-contact 30, portions of the p-contact 30, the p-type region 28, and the active region 26 are removed to expose portions of the n-type region 24 on which the n-contact 32 is formed. The n-and p-

contacts

32 and 30 are electrically isolated from each other by a gap, which may be filled with a dielectric 34 (shown shaded), such as an oxide of silicon or any other suitable material. A plurality of n-contact vias may be formed; the n-and p-

contacts

32 and 30 are not limited to the arrangement illustrated in fig. 2. The n and p contacts may be redistributed to form bond pads with dielectric/metal stacks as is known in the art.

Thick metal pads

36 and 38 are formed on and electrically connected to the n and p contacts. The pad 38 is electrically connected to the n-contact 32. The pad 36 is electrically connected to the p-contact 30.

Pads

36 and 38 are electrically isolated from each other by a gap 40, which gap 40 may be filled with a dielectric material. The gap 40 may be filled with the same dielectric material 12 that separates adjacent LEDs 10 in some embodiments, a different solid material in some embodiments, or air in some embodiments. The gaps 40 are shown shaded. The pad 38 is electrically isolated from the p-contact 30 by the dielectric 34, which dielectric 34 may extend over portions of the p-contact 30.

Pads

36 and 38 may be, for example, gold, copper, an alloy, or any other suitable material formed by electroplating or any other suitable technique.

Pads

36 and 38 are thick enough in some embodiments to support semiconductor structure 22 so that growth substrate 20 can be removed. In this case, the dielectric material 12 provides structural support to the wafer that isolates the LEDs.

A plurality of individual LEDs 10 are formed on a single wafer. In the regions 42 between adjacent LEDs 10, the semiconductor structure is completely removed by etching until the substrate 20, or the semiconductor structure is etched until an electrically insulating layer, as illustrated in fig. 2. As described above with reference to fig. 1, the dielectric material 12 is arranged in the regions 42 between the LEDs 10. The dielectric material 12 may mechanically support and/or protect the sides of the LED10 during later processing, such as dicing. The dielectric material 12 may also be formed to prevent or reduce the amount of light escaping from the sides of the LED 10. The dielectric material 12 may be, for example, epoxy or any other suitable material, and may be formed by any suitable technique including overmolding, spin-on coating, or any other deposition technique. Dielectric material 12 may be formed such that it extends over the bottom of pads 36 and 38 (not shown in fig. 2). Excess material above

pads

36 and 38 may be removed by any suitable technique, such as, for example, bead blasting. In some embodiments, removing excess dielectric material 12 results in a planar surface that includes the bottom surface of dielectric material 12, the bottom surface of material 40, and the bottom surfaces of

pads

36 and 38.

Pads

36 and 38 may then be extended by patterning and depositing extensions 36A and 38A, with extensions 36A and 38A extending below the level of dielectric material 12, as illustrated in fig. 2.

To form the structure illustrated in fig. 1, the growth substrate is removed from the wafer of LEDs 10. The growth substrate may be removed, for example, by laser melting, etching, mechanical techniques (e.g., grinding), or any other suitable technique. The semiconductor structure 22 of the LED10 may be thinned after the growth substrate is removed, and/or the exposed top surface may be roughened, textured, or patterned, for example, to improve light extraction from the LED 10.

The wavelength converting layer 14 is connected to the surface of the LED10 exposed by removing the growth substrate. For example, wavelength-converting layer 14 may be laminated over LED 10.

The wavelength conversion layer 14 may be formed separately from the LED 10. The wavelength converting layer absorbs light emitted by the LED and emits light of one or more different wavelengths. The unconverted light emitted by the LED is typically part of the final spectrum of light extracted from the structure, although it need not be. Examples of common combinations include a blue emitting LED combined with a yellow emitting wavelength converting material, a blue emitting LED combined with green and red emitting wavelength converting materials, a UV emitting LED combined with blue and yellow emitting wavelength converting materials, and a UV emitting LED combined with blue, green and red emitting wavelength converting materials. Wavelength converting materials emitting light of other colors may be added to tailor the spectrum of light emitted from the structure.

The wavelength converting layer is a suitable transparent material such as a silicone or resin loaded with one or more wavelength converting materials such as conventional phosphors, organic phosphors, quantum dots, organic semiconductors, II-VI or III-V semiconductor quantum dots or nanocrystals, dyes, polymers or other materials that emit light. Although the following description refers to phosphors in silicone, any suitable wavelength conversion material or materials and any suitable transparent material may be used. Non-wavelength converting materials may be added to the wavelength converting film, for example to cause scattering or to alter the refractive index of the film.

The wavelength conversion layer may be formed on a roll (roll) of the support film. The support film may, for example, be a commercially available polymer, such as tetrafluoroethylene (ETFE) foil in any suitable size. To form the wavelength conversion layer, phosphor powder is mixed with silicone resinOr other suitable binder, to form a suspension, and the suspension is sprayed or otherwise deposited onto the support film in a continuous process to a predetermined thickness (assuming a continuous dispensing roll). In one embodiment, a YAG phosphor (yellow-green) is used. In another embodiment, the phosphor is a mixed red and green phosphor. Any combination of phosphors may be used in conjunction with the LED light to produce any color of light. The density of the phosphor, the thickness of the layer, and the type of phosphor or combination of phosphors is selected such that the light emitted by the combination of the LED die and phosphor(s) has a target white point or other desired color. In one embodiment, the phosphor/silicone layer will be approximately 30-200 microns thick. Other inert inorganic particles may also be included in the suspension, such as light scattering materials (e.g., silica, TiO)₂). The wavelength converting layer may include a plurality of wavelength converting layers in some embodiments, and may include a non-wavelength converting layer in some embodiments.

The suspension is then dried, such as by an infrared lamp or other heat source. The wavelength converting layer may be tested for its color conversion and matched to a particular LED die, generating a range of peak wavelengths.

To laminate the wavelength converting layer over the LED10, the wavelength converting layer may be spread over the LED 10. The wavelength converting layer 14 may be heated to soften it. A hermetic seal may be formed around the periphery of the wafer. A vacuum is created to remove the remaining air between the wavelength-converting layer 14 and the LED 10. Air between the wavelength converting layer 14 and the LED10 may escape through small holes in the wavelength converting layer 14. Air is then allowed to enter the cavity to pressurize the cavity, thereby pressing the wavelength-converting layer 14 onto the LED 10.

Any other suitable technique other than lamination may be used to attach the wavelength-converting layer 14 to the LED 10.

Transparent layer 16 is then attached to wavelength converting layer 14. Transparent layer 16 may be, for example, a pre-formed glass wafer attached to wavelength conversion layer 14 by a suitable adhesive, such as silicone.

The wafer illustrated in fig. 1 may then be diced into individual LEDs 10 or groups of LEDs 10. In fig. 3, the structure of fig. 1 is arranged in a dicing frame 44. In the dicing lane 46 indicated by the dotted line, three different materials having different mechanical properties must be diced: a transparent material 16, which is typically glass and is hard and brittle; a wavelength converting layer 14, which is typically a silicone-based laminate layer with phosphor particles dispersed in the middle and is soft and nearly tacky, but tough; and a dielectric material 12 between the LEDs 10, which is typically an epoxy with a silica particle filler and is brittle.

Transparent layer 16 is typically the thickest part of the wafer. The dielectric material 12 may be at least 30 microns thick in some embodiments, and no more than 60 microns thick in some embodiments; wavelength-converting layer 14 may be at least 50 microns thick in some embodiments, and no more than 100 microns thick in some embodiments; transparent material 16 may be at least 100 microns thick in some embodiments, and no more than 300 microns thick in some embodiments.

Resin-bonded diamond grit blades are commonly used to cut bare glass wafers on mechanical saws. Mechanical blade slitting relies on abrasion. The blades are formed by using different types of bonding materials to hold diamond grit of specified sizes together. During dicing, the newly exposed tips of the diamond particles are continuously scratched against the wafer. During sawing, the wafer wears on the blade. As the diamond tip becomes dull, the diamond bit falls off the blade and a new bit appears. The abrasive chips produced by sawing are carried away into pockets formed by the diamonds falling from the blade. The glass is resistant to abrasion so the exposed diamond tips quickly dull. Accordingly, conventionally, glass is sawed using a blade having a soft bonding material (e.g., resin) such that most of the outer diamonds are easily dropped, thereby exposing new diamonds in order to maintain the cutting ability of the blade. When sawing abrasion resistant materials such as glass, hard bonding materials such as metals can generate enough heat to melt the blade.

The thinnest resin bonded blade possible is 50-100 microns wide due to blade manufacturing limitations, resulting in a kerf width of 55-110 microns when such blades are used on a wafer. The metal bonded blade can be made 15-20 microns wide, resulting in a kerf width of 20-25 microns. Each wafer illustrated in fig. 1 includes an epitaxially grown semiconductor wafer. Such wafers are expensive to manufacture. Accordingly, the kerf width is kept as narrow as possible to reduce the waste of expensive epitaxial material. Conventional resin bonded blades are not preferred for dicing the wafer illustrated in fig. 1 because the kerf width wastes a large amount of epitaxial material, which may increase the cost of manufacturing the LEDs.

In an embodiment of the present invention, a thin, metal-bonded diamond grit dicing blade is used to dice a wafer including a wavelength converting layer, such as the wafer illustrated in fig. 1. Wavelength converting particles (typically phosphor particles) in the wavelength converting layer and optional particles (e.g. SiO) in the dielectric material 12₂Filler) produces a self-sharpening effect on the blade, which maintains the level of wear. The particles are hard, numerous (e.g., 50 to 60% of the volume of

layers

14 and 12 in some embodiments), and large (e.g., 10 to 50 microns in diameter in some embodiments). When the metal bonded slitting blade impacts these particles at process speeds (in some embodiments, in the range of 30,000 to 50,000 rpm), the opposing abrasion on the blade by the wafer causes substantially uniform wear on the outer layer of the blade. The impact is strong enough that the hard metal bond on the blade erodes, allowing new diamond grit to continue to appear to the surface of the blade, similar to dressing preparation. The newly emerging blade cutting surface as a result of self-sharpening by the above-described particles prevents diamond passivation and enables the glass layer 16 to be cut with a metal-bonded blade.

Thin, metal bonded diamond grit dicing blades such as those conventionally used for Si wafer dicing may be used in embodiments of the present invention. The particular blade used may depend on the particle loading level and the variation in particle size of the above-described blade self-sharpening particles in the wafer, which may be determined by the intended application of the LEDs. The wafer is placed on the dicing frame 44 with the LEDs 10 facing up for alignment, as illustrated in fig. 3. Because wafers can wear the blade, dicing can include blade exposure inspection steps at periodic intervals, for example, after a given number of streets 46 have been cut. The height and/or depth of cut of the blade may be adjusted during the blade exposure inspection step to cause wear on the blade.

In some embodiments, the wafer illustrated in fig. 1 is diced in more than one dicing step. In a first cutting step, a first portion of the thickness of the wafer is cut. In the second cutting step, the remaining thickness of the wafer is cut.

Surface lift-off laser scribing and mechanical sawing with a thin, metal bonded dicing blade are used to dice layers 12 and 14. Both techniques can be adapted for narrow kerf widths, e.g. less than 25 microns. The layer 16 may be singulated by mechanical sawing with a wide, resin bonded blade or using subsurface laser scribing and die breaking. Different permutations of these slicing techniques are described below.

Fig. 4A and 4B illustrate sawing with a wide blade in combination with sawing with a narrow blade. In fig. 4A, the wafer illustrated in fig. 1 is disposed on the dicing frame 44 with the transparent layer 16 facing upward. A wide, resin bonded blade is used to saw through portions of the thickness of the clear layer 16 and wavelength converting layer 14. The sawing process illustrated in fig. 4A produces wide openings 48 that expose the remaining thickness of the wavelength converting layer 14 and the material 12 under the wavelength converting layer 14. In fig. 4B, a thin metal bonded blade is used to saw through the remaining thickness of wavelength converting layer 14 and material 12 to complete the dicing. The sawing in fig. 4B starts at the bottom of the cut 48 opened in fig. 4A. The resulting opening 50 in fig. 4B is narrower than the resulting opening 48 in fig. 4A.

Fig. 5A and 5B illustrate sawing with a wide blade in combination with laser lift-off. In fig. 5A, the wafer illustrated in fig. 1 is disposed on the dicing frame 44 with the transparent layer 16 facing upward. A wide, resin bonded blade is used to saw through portions of the thickness of the clear layer 16 and wavelength converting layer 14. The sawing process illustrated in fig. 5A produces wide openings 52 that expose the remaining thickness of the wavelength converting layer 14 and the material 12 under the wavelength converting layer 14. In FIG. 5B, the remaining thickness of wavelength-converting layer 14 and material 12 are melted using a lift-off laser to complete the dicing. The peeling in fig. 5B begins at the bottom of the incision 52 opened in fig. 5A. The resulting opening 54 in fig. 5B is narrower than the resulting opening 52 in fig. 5A.

Fig. 6A and 6B illustrate sawing with a narrow blade combined with sawing with a wide blade, wherein the wafer is flipped between the two sawing processes. In fig. 6A, the wafer illustrated in fig. 1 is arranged on a dicing frame 44 with the LEDs 10 facing upward. A narrow, metal bonded blade is used to saw through portions of the thickness of material 12 and wavelength converting layer 14 between LEDs 10. The sawing process illustrated in fig. 6A produces narrow openings 56 that expose the remaining thickness of wavelength-converting layer 14 and transparent layer 16. In fig. 6B, the wafer is flipped and placed on the dicing frame 44 with the transparent layer 16 facing upward. A wide resin bonded blade is used to saw through the remaining thickness of the wavelength converting layer 14 and the transparent layer 16 to complete the dicing. The opening 58 created in fig. 6B is wider than the opening 56 created in fig. 6A.

Fig. 7A and 7B illustrate laser lift-off in combination with sawing with a wide blade, where the wafer is flipped between the two processes. In fig. 7A, the wafer illustrated in fig. 1 is arranged on a dicing frame 44 with the LEDs 10 facing upward. Laser lift-off is used to fuse through all or part of the thickness of material 12 and wavelength converting layer 14 between LEDs 10. The lift-off process illustrated in fig. 7A results in a narrow opening 60 that exposes the remaining thickness of wavelength-converting layer 14 and transparent layer 16. In fig. 7B, the wafer is flipped and placed on the dicing frame 44 with the transparent layer 16 facing upward. A wide resin bonded blade is used to saw through the remaining thickness of the wavelength converting layer 14 and the transparent layer 16 to complete the dicing. The resulting opening 62 in fig. 7B is narrower than the resulting opening 60 in fig. 7A.

In the embodiment depicted in fig. 4A, 4B, 5A, 5B, 6A, 6B, 7A and 7B, the transparent layer 16 is cut with a wide, resin-bonded blade. A wide, resin bonded blade produces a large incision. Accordingly, after dicing, in some embodiments, transparent layer 16 is narrower (in a plane parallel to the top surface of LED 10) than the LED10 or grouping of LEDs 10 to which transparent layer 16 is attached. For example, as illustrated in fig. 6B, the width 600 of the transparent layer 16 may be less than the width 602 of the LED 10. In other words, the transparent layer 16 has a smaller lateral extent than the LED10 in a plane parallel to the top surface of the LED 10. Transparent layer 16 may have a smaller lateral extent than LED10 in more than one plane parallel to the top surface of LED 10.

Fig. 8A and 8B illustrate laser lift-off combined with laser scribing and breaking, where the wafer is flipped between the two processes. In fig. 8A, the wafer illustrated in fig. 1 is arranged on a dicing frame 44 with the LEDs 10 facing upward. Laser lift-off is used to fuse through the thickness of material 12 and wavelength-converting layer 14 between LEDs 10. In fig. 8B, the wafer is flipped and placed on the dicing frame 44 with the transparent layer 16 facing upward. An under-surface scribe laser is used to create mechanical damage below the surface of transparent layer 16. Finally, the LED10 in the region 66 (not shown) is separated using a die-breaker to complete the singulation.

Fig. 9A and 9B illustrate sawing with a narrow, metal bond blade combined with laser scribing and breaking, where the wafer is flipped between the two processes. In fig. 9A, the wafer illustrated in fig. 1 is arranged on a dicing frame 44 with the LEDs 10 facing upward. A narrow, metal bonded blade is used to saw through the material 12 and wavelength converting layer 14 between the LEDs 10. The sawing process illustrated in fig. 9A produces narrow openings 68 that expose transparent layer 16. In fig. 9B, the wafer is flipped and disposed on the dicing frame 44 with the transparent layer 16 facing upward. An under-surface scribe laser is used to create mechanical damage below the surface of transparent layer 16. Finally, the LEDs 10 in the region 70 are separated using a die-breaker to complete the singulation.

Having described the invention in detail, those skilled in the art will appreciate that, given the present disclosure, modifications may be made to the invention without departing from the spirit of the inventive concepts described herein. Therefore, the scope of the invention is not intended to be limited to the specific embodiments shown and described.

Claims

1. A method of dicing a wafer of light emitting devices, the wafer comprising:

a glass layer;

a light emitting device layer comprising a plurality of light emitting devices separated by a dielectric such that the dielectric is disposed between the plurality of light emitting devices; and

a single wavelength conversion layer disposed between the glass layer and the light emitting device layer; the method comprises the following steps:

sawing the dielectric between the single wavelength converting layer and the light emitting devices in the light emitting device layer in a dielectric region with a metal bonded diamond grit blade,

maintaining a level of wear of the metal-bonded diamond grit blade by the particles in the single wavelength conversion layer, an

Sawing the glass layer with a metal bonded diamond grit blade after maintaining the level of abrasion.

2. The method of claim 1, wherein:

each light emitting device includes a ill-nitride light emitting layer disposed between an n-type region and a p-type region;

the dielectric includes particles disposed in an epoxy resin; and is

The wavelength conversion layer includes a phosphor disposed in a silicone.

3. The method of claim 1, wherein the glass layer is thicker than the light emitting device layer.

4. A method of dicing a wafer of light emitting devices, the wafer comprising:

a transparent layer;

a single wavelength conversion layer disposed between the transparent layer and the light emitting device layer; the method comprises the following steps:

a first portion of the thickness of the wafer is cut in a first cutting process using a metal bonded diamond grit blade,

The remaining thickness of the wafer is cut in a second cutting process using a metal bonded diamond grit blade maintained at an abrasive level.

5. The method of claim 4, wherein the second cutting process begins in the incision opened in the first cutting process.